Psychometric validation of the Spanish version of the Expanded Prostate Cancer Index Composite-26

Prostate cancer is the most frequently diagnosed non-cutaneous cancer and the fifth cause of death among men in Europe and Spain [1]. Currently, most prostate cancer patients are diagnosed in localized stages [2] and will probably be long-term survivors [3]. Even so, these men may suffer relevant treatment and disease-related side-effects [4, 5], and thus, Patient-Reported Outcome Measures (PROMs) have become relevant endpoints that should be gathered from patients with localized or advanced prostate cancer [6].

The most established disease-specific PROM for these patients is the Expanded Prostate cancer Index Composite (EPIC), one of the instruments with the best properties for evaluating prostate cancer care [7, 8]. It is composed by 50 items, and was developed to expand the scope of the original 20-item University of California, Los Angeles Prostate Cancer Index (UCLA-PCI), with additional items that assess irritating urinary symptoms and the impacts of hormonal therapy [9].

In 2010, a new abbreviated version with 26 items (EPIC-26) was developed by assessing items for elimination through an iterative process. Each domain was correlated with the corresponding scores from the full EPIC-50, and this process was repeated until all psychometric properties reached acceptable levels [9]. The EPIC-26 has been included in the standard set of outcomes recommended by the International Consortium for Health Outcomes Measurement (ICHOM) [6].

Psychometric evaluations of EPIC-26 are available for several country versions (Norway, China, Germany, Italy, and Canada [10‐15]). However, this assessment has not been performed for the Spanish version of the EPIC-26 [16], which was constructed by selecting the corresponding items from the Spanish version of EPIC-50 [17]. The aim of this study is to assess the metric properties of the Spanish version of EPIC-26 in a sample of Spanish men with prostate cancer, in terms of validity, reliability, and sensitivity to change.

Characteristics of patients and the distribution of EPIC scores at 12 months after treatment are summarized in Table 1, separately for those who responded to EPIC-26 (n = 543) and to EPIC-50 (n = 324). In both groups, most participants were categorized at intermediate D’Amico tumoral risk (≈ 40%), and external radiotherapy was the predominant treatment, followed by radical prostatectomy, brachytherapy as monotherapy, and active surveillance. Four patients from the radical prostatectomy group received salvage radiotherapy, and 64 from external radiotherapy group received a boost of brachytherapy. The distribution of the EPIC-26 and EPIC-50 scores at 12 months after treatment shows that the observed range is very similar in all domains to the theoretical range (from 0 to 100), except in the sexual domain. The floor effect was lower than 2% in all domains, but the ceiling effect (percentage of patients with the best outcome) was high in all domains, except for sexual. It was greater than 50% in urinary incontinence and bowel in both versions. Cronbach’s alpha was greater than 0.7 in all domains, except in the urinary irritative/obstructive domain, which was 0.67 for EPIC-26 and 0.63 for EPIC-50. Distribution of EPIC-26 items at 12 months after treatment (absolute and relative frequencies) is presented in Supplementary Table 1.

Results of the confirmatory factor analysis of EPIC-26 with a five-domain structure is represented in Fig. 1, showing acceptable goodness-of-fit indexes: RMSEA = 0.066; CFI and TLI > 0.95. The inter-domains covariances ranged from 0.16 (between urinary incontinence and bowel) to 0.79 (between hormonal and sexual). Most of the standardized regression weights between each item and their domain are greater than 0.6. For instance, the weights between items ‘4b-dysuria’, ‘4c-hematuria’, ‘4d-weak stream’, and ‘4e-frequency’ in the urinary irritative/obstructive domain is 0.61, 0.74, 0.84, and 0.90, respectively. The residuals were lower than 0.5 in all items (the lower the values, the better), except for four items: ‘4b-dysuria’, ‘6d-bloody stools’, ‘12-overall sexual problem’, and ‘13c-depression’.

Supplementary Fig. 1 represents the scatter plots constructed from the subsample of patients who answered the EPIC-50, and the agreement between their scores and those scores obtained by selecting the items that constitute the EPIC-26. The agreement between them was excellent for all domains, with ICCs higher than 0.95.

Table 2 shows the results of the EPIC-26 and the EPIC-50 for the subsample of patients that underwent radical prostatectomy (n = 95 and n = 79, respectively) or external radiotherapy (n = 115 and n = 170, respectively). The patients who underwent radical prostatectomy presented statistically significant changes in urinary incontinence, bowel, and sexual domains, which were of large magnitude (ES > 0.8) in both versions. In those patients treated with external radiotherapy, changes in bowel, sexual, and hormonal domains were statistically significant, and the ES ranged from 0.44 to 0.69 with EPIC-26, and from 0.3 to 1.09 with EPIC-50.

The Spanish version of the EPIC-26 has demonstrated adequate metric properties, similar to those of the original version, with acceptable goodness-of-fit indexes in the confirmatory factor analysis of the five-domain structure and good criterion validity compared to the Spanish EPIC-50. It presented a very low floor effect (< 2%), good reliability in almost all domains (Cronbach’s alpha > 0.7), and responsiveness to detect change after radical prostatectomy or external radiotherapy. However, the ceiling effect was high, especially in urinary incontinence and bowel domains (> 50%).

The Spanish version of the EPIC-26 demonstrated an acceptable fit to the five-domain structure proposed by the original instrument [10]. These results support the construct validity of calculating separate scores for urinary incontinence, urinary irritative/obstructive, bowel, sexual, and hormonal symptoms, similarly to the original study [10] and the German version [13]. The Norwegian [11] and the Canadian version [15] found that a six-domain structure model fits better, splitting up the hormonal domain into two subdomains. Consistently with results of the CFA obtained in the Norwegian [11], the German [13], and the Canadian version [15], we also found it difficult to make the model converge, which was solved by dichotomization of items (no problems vs problems), and the exclusion of the item ‘9-Erection not firm’.

Criterion validity results showed an excellent agreement (ICC > 0.90 [22]) between all EPIC-26 domains and those from the original EPIC-50. A similar evaluation has only been performed in the original American version study [10]. Despite estimating different validity parameters, the results for criterion validity of EPIC-26 compared to EPIC-50 domains were almost identical: the correlation r ranged from 0.96 to 0.97 in the original version [10], and the ICC ranged from 0.96 to 0.98 in our study. This indicates that regardless of the EPIC version a patient completed, the domain scores obtained would be practically the same. Hence, considering its low burden of administration, the abbreviated version with 26 items is more appropriate.

The Spanish version of the EPIC-26 presented a floor effect near zero, in line with those observed in other studies of EPIC-26 versions [11, 15], and meets the recommended quality criteria for floor effect (lower than 15%) [26]. In contrast, bowel and urinary incontinence domains exhibited high ceiling effects (% of patients with the best outcomes), similarly to the studies of the American [10], the Norwegian [11], and the Canadian versions [15]. Nonetheless, these percentages in our study were higher (66.1% and 54.1%, respectively) than in the previous publications. For instance, in the bowel domain: 34% in the American version, and close to 62% in other versions [11, 15]. Considering these high percentages of patients reporting no problems in bowel items, as well as in urinary incontinence, the evaluation of the EPIC-26 validity in those patients with slight or mild symptoms in these domains merits further research. It is important to consider that the EPIC-26 was developed to measure the impact of a broad spectrum of treatments differing in side-effects [9, 10], which explains the variation of the ceiling effect across different samples according to treatments applied: for instance, urinary incontinence is common after surgery, and bowel discomfort after radiotherapy.

Reliability results achieved the acceptable threshold for internal consistency (Cronbach’s alpha > 0.7) [20] in most EPIC-26 domains. They were very similar to those obtained by the EPIC-50, indicating that the EPIC-26 contains enough items to measure domains without internal consistency penalization. Similarly to the Chinese [12], German [13], and Italian version [14], the urinary irritative/obstructive domain is the only one that showed poor internal consistency in our study (Cronbach’s alpha = 0.67). This may be explained by the item ‘4c-Hematuria’, with an extreme percentage of patients reporting no problems (98.5%). In fact, Cronbach’s alpha of this domain was 0.71 when excluding this item (data not shown). This is in line with the Italian version study [14], which highlights the need of further research to identify more reliable new urinary irritative/obstructive items. However, the most recently developed prostate cancer-specific PROM, the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-Prostate (EORTC QLQ-PR25), does not measure urinary irritative/obstructive symptoms [27].

The Spanish version of the EPIC-26 is able to detect changes between before and 12 months after treatment as hypothesized a priori, according to clinically known side-effects of surgery or external radiotherapy for prostate cancer [4, 24, 25]. On the one hand, patients who underwent external radiotherapy presented statistically significant deteriorations of moderate magnitude in bowel, sexual, and hormonal domains (ES ranging from 0.44 to 0.69). On the other hand, patients who underwent radical prostatectomy presented statistically significant deteriorations of large magnitude in urinary incontinence (ES = 2.25) and sexual domains (ES = 0.95). To the best of our knowledge, our study is the first one that assesses EPIC-26 responsiveness for surgery and external radiotherapy, since this property has not been evaluated in the original American version nor in other studies of EPIC-26 versions. Although the German study [13] also evaluated responsiveness, they considered the whole sample (84% of patients underwent radical prostatectomy and 16% had other treatments), even though side-effects differ according to the treatments applied. Our results of patients who underwent radical prostatectomy are in line with the deterioration of large magnitude obtained in the German study (ES = 1.22 for urinary incontinence and ES = 1.15 for sexual domains) [13].

The main limitation of this study is that it includes patients with localized or locally advanced prostate cancer, mostly treated with external radiotherapy or radical prostatectomy, which limits the generalizability of the results to patients with metastatic disease or to patients who underwent other treatments. However, the sample is heterogeneous enough to represent most tumoral stages, as well as the most currently established treatments for non-metastatic disease. Furthermore, test–retest reproducibility could not be studied, since the time elapsed between administrations was too long to assume the patients’ stability. Nevertheless, good results of internal consistency support the reliability of the EPIC-26.

In conclusion, this study represents the first evaluation of the psychometric properties of the Spanish version of the EPIC-26, which can be considered a reliable and valid instrument to analyze the impact of different treatments in patients with localized or locally advanced prostate cancer. The good responsiveness of EPIC-26 to detect changes after treatment supports its usefulness for the clinical decision-making process in these patients. In addition, the low burden of administration makes the EPIC-26 a practical tool for its routine use in clinical practice and in international multicentric studies, and it could facilitate benchmarking among registries.